Reflective intelligence surface technology for future wireless networks

Reconfigurable intelligent surfaces (RISs) have witnessed significant attention due to their potential to improve the efficiency and coverage of wireless networks. RIS acts as a smart mirror, which reconfigures the wireless propagation environment by tuning the incoming waveform’s phase shift, amplitude, and polarization. To fully realize the capabilities of RIS, the phase shifts should be efficiently optimized. Researchers have considered optimization-based techniques to tackle the phase shift optimization problem. However, such methods are complex in nature and are difficult to realize for large-scale systems. To this end, deep reinforcement learning (DRL) has emerged as a robust and powerful approach for optimizing wireless communication systems. DRL learns from interacting with the environment without needing a labeled dataset, enabling adapting to the dynamic changes in the communication environment. In this work, we develop DRL frameworks to optimize full-duplex (FD) RIS-assisted communication systems. FD communications are envisioned as one of the essential technologies for future wireless communications. Incorporating RIS into FD systems can efficiently establish a reliable communication system and resolve the co-channel interference issue of FD systems. To this end, this work first proposes a low-complexity DRL algorithm to optimize the RIS phase shifts of a half-duplex (HD)-FD RIS-assisted communication system. The proposed algorithm is the first of its kind, which tackles the optimization problem in the FD operating mode. It was shown that the proposed algorithm significantly improves the rate compared to the non-optimized case in both operating modes and reduces the computational complexity compared to the state-of-the-art algorithm in the HD operating mode. Furthermore, the deployment of distributed RISs is also investigated in this thesis. In particular, the preference of deploying single or distributed RIS schemes is studied based on the links’ quality considering three practical scenarios. The sum-rate maximization problem is considered subject to transmit beamformers and RIS phase shifts of a FD RIS-assisted communication system. To address the optimization problem, a two-step solution is proposed. First, a closed-form solution is derived to optimize the beamformers. Second, a DRL algorithm is proposed to optimize the RIS phase shifts. The proposed solution was shown to efficiently outperform the conventional beamformers approximation and improve the sum rate compared to the non-optimized RIS phase shifts. Finally, this work considers a DRL approach for optimizing the discrete phase shifts of FD distributed RIS-assisted system. The discrete phase shifts are considered to offer a feasible solution, since the continuous phase shifts are infeasible to implement due to hardware limitations. A deep Q-learning algorithm is developed to optimize the RIS phase shifts, along with two mathematical beamformers derivations (i.e., closed-form and approximate). The performance of the proposed algorithm is further assessed through extensive simulations by considering two scenarios: the presence of the line-of-sight (LoS) link and when it is blocked. It was shown that the proposed algorithm achieves promising results compared to the ideal approach (the continuous baseline), which guarantees a near-optimal performance. The complexity analysis for all proposed algorithms and simulation results are provided to support these findings.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.